- Oct 31, 1999
- 30,699
- 1
- 0
OK, maybe those who run pProxies can help me with this one...
I used to run the client with frequent buffer-level checking, which kept all work units flushed and kept the in-buffer full. A couple of times, people whose pProxy I was using mentioned that the client was making connection every 30 seconds even though it didn't need to... "going through the motions." I have since disabled additional buffer-level checking so as to not bog down Mika's, but I was wondering if there's a happy medium where the client will send each work unit as it's completed, but won't harass the pProxy every 30 seconds. Looking at the client's menu, I see the following:
distributed.net client configuration: Buffer and Buffer Update Options
--------------------------------------------------------------------------
Additional buffer-level checking:
The following options are extensions to normal threshold management and are
not usually necessary:
0) no additional buffer-level checking. (default)
1) ensure that there is always work available.
2) ensure that all completed work is kept flushed.
3) both 1) and 2). (implied if 'Dialup detection options' are enabled)
4) update on per-project buffer exhaustion.
Options 1, 2 and 3 will cause the client to frequently check buffers levels.
(Frequency/interval is determined by the 'Buffer-level check interval' option)
You might want to use them if you have a single computer with a network
connection "feeding" other clients via a common set of buffers, or if you
want to ensure that completed work is flushed immediately.
Option 4 is a hint to the client to work on a single project as long as
possible (updating per-project buffers individually), rather than loop through
all active/enabled projects (one combined update per pass).
Default Setting: 0
Current Setting: 0
New Setting --> 0
It sounds like #2 is what I want, but is the client going to try to connect every 30 seconds? There was also some question that in the incident I mentioned, my client was trying to connect every 30 seconds because I had ordered it to buffer more work units than it was allowed to have, making it want to fetch them but then slap its own hand away, so to speak. Any insights are appreciated!
I used to run the client with frequent buffer-level checking, which kept all work units flushed and kept the in-buffer full. A couple of times, people whose pProxy I was using mentioned that the client was making connection every 30 seconds even though it didn't need to... "going through the motions." I have since disabled additional buffer-level checking so as to not bog down Mika's, but I was wondering if there's a happy medium where the client will send each work unit as it's completed, but won't harass the pProxy every 30 seconds. Looking at the client's menu, I see the following:
distributed.net client configuration: Buffer and Buffer Update Options
--------------------------------------------------------------------------
Additional buffer-level checking:
The following options are extensions to normal threshold management and are
not usually necessary:
0) no additional buffer-level checking. (default)
1) ensure that there is always work available.
2) ensure that all completed work is kept flushed.
3) both 1) and 2). (implied if 'Dialup detection options' are enabled)
4) update on per-project buffer exhaustion.
Options 1, 2 and 3 will cause the client to frequently check buffers levels.
(Frequency/interval is determined by the 'Buffer-level check interval' option)
You might want to use them if you have a single computer with a network
connection "feeding" other clients via a common set of buffers, or if you
want to ensure that completed work is flushed immediately.
Option 4 is a hint to the client to work on a single project as long as
possible (updating per-project buffers individually), rather than loop through
all active/enabled projects (one combined update per pass).
Default Setting: 0
Current Setting: 0
New Setting --> 0
It sounds like #2 is what I want, but is the client going to try to connect every 30 seconds? There was also some question that in the incident I mentioned, my client was trying to connect every 30 seconds because I had ordered it to buffer more work units than it was allowed to have, making it want to fetch them but then slap its own hand away, so to speak. Any insights are appreciated!
Facebook
Twitter